Study participants characteristics:
Median values of the study participants were age 38 years, body mass index (BMI) 26.4, and waist circumference 91.7 cm. The median serum copper level was 113.6 µg/L, median selenium level of 126.1 µg/L and the median serum zinc level was 80.7 µg/L (Table 1). Other demographics, blood pressure and laboratory parameters relevant to this study are shown in Table 1 (see Supplementary Files).
Unadjusted odds ratios of relevant variables and the trace elements with hypertension:
Lowest and highest quantiles for each of the trace elements were screened for their association with hypertension. The lowest quantile of serum selenium and serum copper negatively correlated with hypertension, while the highest quantile of serum selenium was positively correlated with hypertension (Table 2 in the Supplementary Files).
Unadjusted odds ratios were calculated for each of the relevant variables for hypertension causation and the serum trace elements (Table 2). The results of the unadjusted odds ratios with 95% confidence intervals (CI) are shown in Figure 1. The median serum selenium among hypertensives was 126.1 µg/L, but only 9.3% of the participants (n=625) had serum selenium higher than the normal range (70 – 150 µg/L) . Studies suggested that at or around 120µg/L, selenium values are optimal for distribution and transport . In this analysis, at high normal range of serum selenium value at 120 µg/L or more, the serum selenium level was significantly associated with hypertension (unadjusted OR = 1.78, 95% CI= 1.58 – 2.022). Interestingly, serum selenium higher than 150 µg/L strengthened the association with hypertension (unadjusted OR= 1.84, 95% CI = 1.55-2.18). Estimated probabilities comparisons of serum trace elements selenium, zinc and copper are shown in Figure 2. Increased probability for hypertension was noted only in case of high serum selenium levels.
Predicted probabilities for the serum trace elements with hypertension
The predicted probabilities of serum selenium, serum copper and serum zinc as an output from logistic regression analysis revealed no significant association of serum copper and serum zinc levels with hypertension (Figure 2). However, there was a significant positive association of high serum selenium with high blood pressure (Table 2, Figures 1 and 2). These initial findings of association concurred with subsequent analyses. Although lower than normal serum zinc levels showed odds ratio of 1.4 (1.01 – 1.51) (Figure 1), the subsequent polynomial regression procedure with adjustments for confounding factors on low serum zinc did not show strong association with hypertension (Figure 1 and Table 3 in the Supplementary Files) and was not analyzed further.
Confounding factors adjustments with polynomial regression models
Confounding factors considered for multinomial logistic regression model were smoker, gender (male), BMI > 24, borderline high cholesterol, high serum cholesterol, high waist circumference for males and females, daily sodium and cholesterol intake and taking anti-hypertensive medications (Figure 1). Among the different races, being in the black population of non-Hispanic origin was considered as a confounding factor in the analysis, but pregnancy was not (Figure 1).
After adjusting for confounders, serum selenium levels of 120 µg/L or higher (reference level 75 -150 µg/L) were significantly associated with high blood pressure (OR = 1.56(1.38 – 1.76)). Also, at serum selenium greater than 150 µg/L, the association with high blood pressure strengthened (OR = 1.91 (1.54 – 2.36)) with hypertension after adjusting for confounding factors (Table 3). The adjusted odds ratios of hypertension with serum selenium at highest quantile versus lowest quantile were significant at 1.34 (95% CI= 1.17 -1.53) and 0.63(95% CI= 0.55-0.72) respectively.
When confounding factors were taken into account, serum copper and zinc were no longer significantly associated with hypertension, with either higher or lower than normal values or at their highest and lowest quantiles (Table 3). The adjusted odds ratios of hypertension with serum copper at highest versus lowest quantiles were 1.03(95% CI=0.9-1.19) and 1.05(95% CI=0.91-1.2) respectively. For serum zinc, the adjusted odds ratios of hypertension at highest versus lowest quantiles were 0.99(95% CI = 0.87-1.13) and 1.15(95% CI=1.00-1.31) respectively.
Since the association of increments in serum selenium was not linear at all serum selenium levels and the association with hypertension increased from values at 150ug/L compared to 120ug/L (Figure 2 and Table 3), quantile regression models were also analyzed for serum selenium, systolic blood pressure, diastolic blood pressure, age of the participants and total serum cholesterol levels (Figures 3 and 4). Looking closer at the effect of serum selenium on systolic and diastolic blood pressure, the results of quantile regression revealed stronger increments in associations of diastolic blood pressure with the higher quantiles of selenium compared to systolic blood pressure, where the strength of association remained mostly at the same level in all the quantiles of serum selenium (Figure 3). Also, the quantile regression models for serum levels of selenium showed a positive association on all the quantiles of serum selenium with increments of the participant age, and a stable but stronger association with total serum cholesterol levels persisted at all quantiles of serum selenium (Figure 4), as some studies suggest .